This document presents an approach using the maximum likelihood formulation to estimate vector velocities in real-time by a network of Doppler radars. Relationships between the estimated vector velocity, the statistics of the measured signals, the characteristics of the observing geometry, and the hardware and signal processing parameters is derived. Metrics to gauge the quality of vector velocity retrievals are presented, and their utilization for network design and operation is provided. The thesis concludes with a software architecture for real-time implementation of the vector velocity estimation and its demonstration within the framework of the CASA IP1 four node radar network.